Everything Jeremy Achin said on any show that made the record, most notable first. Each card names its show and opens the statement there.
Achin: Compliance is used to defend outdated, inaccurate regression models
“A word people use as a last resort to defend the status quo after they realize that their hundred variable regression model is an arbitrary representation of reality that is less accurate, robust, and interpretable than the modern, ah, alternatives.”
Achin: Automated tools and education will solve the data scientist shortage
“And, ah, the shortage of data scientists will be solved by a combination of pragmatic education, practical education, and, ah, tools that automate the modeling process and automate data science to levels which are, right now, probably most of us in the room th…”
Achin: Complex machine learning models can be interpreted using modern techniques
“I would argue, though, that even very complicated models can be interpreted using modern techniques.”
Achin: 'Black box' label reflects fear of new technology
“It's a phrase people use when they're scared of technology they don't understand and want to keep doing the same thing they've been doing for the last 20 years.”
Achin: DataRobot employees held Kaggle's top rank almost continuously for three years
“Basically, over the last three years, somebody from our company has been number, ranked number one almost the entire time.”
Achin penalized job candidates for choosing the wrong modeling textbook
“I love this book so much that I would use it in interviews, so I, when somebody would come in for an interview, I would ask them, if you had to go on a desert island And do modeling consulting for the rest of your life, and you can bring one book, what would i…”
Achin: Non-data scientist advisor placed 5th in Kaggle competition using DataRobot
“One of our advisors, who's not a data scientist, knows nothing about insurance he came in fifth place. And this is out of six hundred-something people including thirty-something Liberty Mutual employees that have competed.”
Achin: Gradient boosted trees perform best over 40% of the time
“Gradient boosted trees tends to be you know, the, one of the top algorithms, like 40 plus percent of the time.”
Achin: Large companies still employ armies of people manually building regression models
“But there's still armies of people mostly in big companies you know, building regression models manually, taking months to do that.”